The present invention relates to a method for getting a system dump that collects information required to investigate a cause of a failure when the failure of a computer system occurs.
In the event that a failure occurs during operation of the computer system, it is necessary to obtain content of virtual storage, which has been utilized at that time by the computer system, in order to investigate its cause. Hereinafter, the information is referred to as “dump information”.
On the other hand, a workload executed by a computer system, and quantity of transacted data, are increasing year by year. In order to cope with the situation, sizes of virtual storage and main storage, which can be managed by the computer system, are also becoming larger. This results in an increase in quantity of dump information, which should be gotten in the event of a failure, causing time for getting the dump information to be lengthened. Therefore, time required to restart the system after the occurrence of the failure becomes longer.
Methods for avoiding such a delay in system restart are described in Japanese Patent Laid-Open No. Hei 7-234808 and Japanese Patent Laid-Open No. Hei 10-333944.
In the Japanese Patent Laid-Open No. Hei 7-234808, a method for getting a dump of a computer system having duplexed main storage is disclosed. In the Japanese Patent Laid-Open No. Hei 7-234808, a method for preventing a delay in system restart is disclosed. In this case, the delay is prevented in the following manner: in the event that an abnormal condition of a system occurs, memory information on one side is gotten as dump information, and the system is restarted using a memory on the other side.
In the Japanese Patent Laid-Open No. Hei 10-333944, a method for preventing a time delay in restarting a system is disclosed as follows. First of all, a dump for a memory area, in which a core portion of an operating system is loaded at the time of restart, is acquired before starting operation for restarting the system. After that, the following programs are executed in parallel: a program for restarting the system while dumping a memory area, which will be used, before using the memory area, if the memory area has not been dumped; and a program for successively dumping areas that have not been dumped.
As described above, in the prior art, the methods for outputting information, which is stored on the main storage, as dump information is disclosed. However, concerning a computer system that adopts virtual storage management, when an abnormal condition of a system occurs, data, which has been paged out to an auxiliary storage, should also be obtained as dump information.
The computer system, which adopts the virtual storage management, may output (page out) a part of virtual storage to an external storage in order to perform concurrent processing in large quantity in the main storage having a limited size, or may input (page in) an area of the virtual storage to the main storage when referring or updating the area of the virtual storage. In the prior art described above, there is no description of a method for shortening time required to restart a system while enabling acquisition of such dump information that is not allocated on main storage.
To be more specific, in the prior art described above, when a failure occurs in a computer system, if it is required to acquire information on the whole virtual storage, restart of the computer system is not permitted until information, which has been paged out to the auxiliary storage, is acquired as dump information. In addition, in the prior art described above, it is necessary to add an external storage for storing the dump information.